Follow these steps to create a visual recognition model in Watson Studio. For this tutorial, you can use the data files in this github repository.

Create the model

  1. From the Projects page in Watson Studio, add a new project.
  2. When prompted, select the Visual Recognition project type.
  3. If you have no instances of the Visual Recognition service, a new instance will be created. If you have existing instances, you have a choice: Click the Existing option to associate an existing instance with the project. Click the New option to create another new instance.
  4. Click Add to project, and then select Visual Recognition model.
  5. If a Visual Recognition service instance is not already associated with your project, follow the prompts to create a new service or associate an existing service. Note: A Visual Recognition service instance can be associated with only one project at a time.
  6. Enter a name for your model.
  7. Add .zip files to your model from the data panel. Don’t add all of the sample images at this point, so you’ll have additional images for retraining the model later. (If the data panel isn’t open, you can open it by clicking the Find and add data icon.)
  8. (Optional) Adjust your classes, for example: remove or rename classes, remove images from a class, or move images from one class to another.
  9. Click Train Model.
  10. Wait for the training process to finish. You can leave the Visual Recognition model builder while the process runs. You can monitor the status of the training process by looking at the status of the model on the Assets page of your project.

Test the model

  1. When you see the message “Training successful. Your model training is successful. click here to view and test your model.”, click the link to test your new model.
  2. Test individual image files by dragging them from your computer onto the test area. Or test all the images in a .zip file from the the data panel. (If the data panel isn’t open, you can open it by clicking the Find and add data icon.)

Retrain the model

  1. From the model test tab, click Edit and Retrain.
  2. Edit your model:
    1. Add more classes with new training images to your model.
    2. Rename classes to eliminate any invalid characters.
    3. Add negative images to the Negative class.
    4. Reclassify any images.
  3. Click Train Model.

Test the model again

  1. When you see the message “Training successful. Your model training is successful. click here to view and test your model.”, click the link to test your new model.
  2. Test individual image files by dragging them from your computer onto the test area. Or test all the images in a .zip file from the the data panel. (If the data panel isn’t open, you can open it by clicking the Find and add data icon.)

Additional tutorial

Try this tutorial: Survey wildfire-damaged neighborhoods to identify burned homes and intact homes

Join The Discussion

Your email address will not be published. Required fields are marked *